Matthias Max bitflower

LLM Wiki

A pattern for building personal knowledge bases using LLMs.

This is an idea file, it is designed to be copy pasted to your own LLM Agent (e.g. OpenAI Codex, Claude Code, OpenCode / Pi, or etc.). Its goal is to communicate the high level idea, but your agent will build out the specifics in collaboration with you.

The core idea

Most people's experience with LLMs and documents looks like RAG: you upload a collection of files, the LLM retrieves relevant chunks at query time, and generates an answer. This works, but the LLM is rediscovering knowledge from scratch on every question. There's no accumulation. Ask a subtle question that requires synthesizing five documents, and the LLM has to find and piece together the relevant fragments every time. Nothing is built up. NotebookLM, ChatGPT file uploads, and most RAG systems work this way.

description	Start a brainstorming session to create a feature specification
argument-hint	<feature-name>

Feature Specification or Bug Investigation

Start a collaborative session for: $ARGUMENTS

	import { Component, h, Element, Prop, Host, Event, EventEmitter } from '@stencil/core';

	@Component({
	tag: 'animatable-component'
	})
	export class AnimatableComponent {
	@Element() el!: HTMLElement;
	@Prop() keyFrames: Keyframe[]
	@Prop() options: KeyframeAnimationOptions
	@Event() finish: EventEmitter;

	# ---------------------------------------------------------------------------
	#
	# Description: This file holds all my BASH configurations and aliases
	#
	# Sections:
	# 1. Environment Configuration
	# 2. Make Terminal Better (remapping defaults and adding functionality)
	# 3. File and Folder Management
	# 4. Searching
	# 5. Process Management

	#!/usr/bin/ruby
	# encoding: utf-8
	# Grab google web fonts and embed them as base64 data URIs
	# <http://brettterpstra.com/2015/03/14/embedding-google-web-fonts/>
	require 'base64'

	if ARGV.length > 0
	input = ARGV
	elsif STDIN.stat.size > 0
	input = STDIN.read.strip.split(/\n+/)

	/**
	* A generic confirmation for risky actions.
	* Usage: Add attributes: ng-really-message="Are you sure"? ng-really-click="takeAction()" function
	*/
	angular.module('app').directive('ngReallyClick', [function() {
	return {
	restrict: 'A',
	link: function(scope, element, attrs) {
	element.bind('click', function() {
	var message = attrs.ngReallyMessage;

	/* Original is in LESS and can be found here: https://gist.github.com/gefangenimnetz/3ef3e18364edf105c5af */

	@mixin material-shadow($level:1){

	@if $level == 1 {
	box-shadow: 0 1px 3px rgba(0,0,0,0.12), 0 1px 2px rgba(0,0,0,0.24);
	}
	@else if $level == 2 {
	box-shadow: 0 3px 6px rgba(0,0,0,0.16), 0 3px 6px rgba(0,0,0,0.23);
	}

	<?php

	/**
	* UserAgentDetector
	*
	* @author Exadra37
	* @license Free to use at your own risk. Provided as it is, without any warranty that will work for you.
	* @link http://presta-hosting.com
	* @version 1.0.1
	* @since v.1.0.0 - 09/03/2014

	app.filter('trusthtml', ['$sce', function ($sce) {
	return function(t) {
	return $sce.trustAsHtml(t)
	}
	}]);

	img
	// graycity($gray, $opacity)
	@include graycity(0.5, 0.7)

Matthias Max bitflower

LLM Wiki

The core idea

Feature Specification or Bug Investigation

Mode Detection